单词学习：![[d885596f-7313-402c-b918-7ddd2cd7707b.mp3]]

新闻播报：![[6161b321-a3c0-4439-bfc1-da2a559221fb.mp3]]

谷歌Gemini 3发布在即，多线布局追赶GPT-5

简报：

• 谷歌CEO确认其下一代大模型Gemini 3将于2025年发布，旨在追赶竞争对手OpenAI已于同年8月推出的GPT-5。
• 新模型将重点强化智能代理（AI Agent）能力，使其能主动规划、调用工具并执行多步骤任务，实现从“问答助手”到“行动代理”的转变。
• 为扩大AI用户基础，谷歌与印度电信运营商Jio达成合作，向超过5亿用户免费提供Gemini服务；同时将为苹果2026年推出的新版Siri提供定制AI技术。
• 谷歌正通过深度集成生成式AI来推动搜索和云业务增长，并已开始在AI生成内容中测试广告以探索商业化路径。
• Alphabet旗下自动驾驶公司Waymo也正加速国际化布局，计划将AI驱动的出行服务拓展至伦敦、东京等全球主要城市。

相关链接：

• Gemini 3明年登场！谷歌全力追赶GPT-5，6.5亿用户成AI反攻关键筹码
• 谷歌联手印度信实Jio，免费向5亿用户推送Gemini AI！人均396美元福利引爆新兴市场
• Siri要翻身了？苹果明年三月推“Apple Intelligence”大改版，竟用上谷歌Gemini！
• 谷歌启动高级搜索广告实验，未来将在其AI生成式内容中展示广告

Google Gemini 3 Release Imminent, Multi-pronged Strategy to Catch Up with GPT-5

Briefing:

• Google's CEO has confirmed that its next-generation large language model, Gemini 3, will be released in 2025, aiming to catch up with rival OpenAI's GPT-5, which was launched in August of the same year. 🚀
• The new model will focus on enhancing its AI Agent capabilities, enabling it to proactively plan, utilize tools, and execute multi-step tasks, thereby transforming from a "Q&A assistant" into an "action agent."
• To expand its AI user base, Google has partnered with Indian telecom operator Jio to provide Gemini services for free to over 500 million users. Additionally, it will supply custom AI technology for Apple's new Siri, set to launch in 2026.
• Google is driving growth in its search and cloud businesses by deeply integrating generative AI, and has begun testing advertisements within AI-generated content to explore commercialization avenues. 💰
• Alphabet's autonomous driving company, Waymo, is also accelerating its international expansion, planning to extend its AI-powered mobility services to major global cities like London and Tokyo. 🌍

Related Links:

• Gemini 3明年登场！谷歌全力追赶GPT-5，6.5亿用户成AI反攻关键筹码
• 谷歌联手印度信实Jio，免费向5亿用户推送Gemini AI！人均396美元福利引爆新兴市场
• Siri要翻身了？苹果明年三月推“Apple Intelligence”大改版，竟用上谷歌Gemini！
• 谷歌启动高级搜索广告实验，未来将在其AI生成式内容中展示广告

苹果AI战略曝光：新Siri将于明年三月发布，集成谷歌Gemini

简报：

• 苹果计划于2026年3月推出代号为“Apple Intelligence”的重大更新，核心是发布经过彻底重构的新一代Siri。
• 新Siri将引入谷歌Gemini大模型的定制版本，该版本将在苹果的私有云服务器上运行，以增强网页理解和实时信息检索能力。
• 苹果CEO蒂姆·库克确认了2026年的发布计划，并表示公司对在AI领域的收购与合作（如与OpenAI）持开放态度。
• 为配合新Siri，苹果还将推出一款介于HomePod与iPad之间、内置高品质扬声器的全新智能家居显示屏设备。
• 完整的AI战略，包括将Apple Intelligence全面集成到iOS 27、macOS 27等新一代操作系统中，将在2026年6月的WWDC上正式公布。

相关链接：

• Siri要翻身了？苹果明年三月推“Apple Intelligence”大改版，竟用上谷歌Gemini！
• 库克:苹果将在人工智能领域寻求更多收购与合作机会

Apple's AI Strategy Unveiled: New Siri to Launch Next March, Integrating Google Gemini 🚀

Brief:

• Apple plans to launch a major update code-named "Apple Intelligence" in March 2026, centered around the release of a thoroughly reconstructed next-generation Siri.
• The new Siri will feature a customized version of Google's Gemini large language model, which will operate on Apple's private cloud servers to enhance web comprehension and real-time information retrieval capabilities.
• Apple CEO Tim Cook confirmed the 2026 launch plan and stated that the company is open to acquisitions and partnerships in the AI domain (such as with OpenAI).
• To complement the new Siri, Apple will also introduce a new smart home display device, positioned between a HomePod and an iPad, featuring high-quality built-in speakers. 🏡
• The full AI strategy, including the comprehensive integration of Apple Intelligence into next-generation operating systems like iOS 27 and macOS 27, will be officially unveiled at WWDC in June 2026. 🤔

Related Links:

• Siri to make a comeback? Apple to launch "Apple Intelligence" major revamp next March, surprisingly using Google Gemini!
• Cook: Apple will seek more acquisition and partnership opportunities in the field of artificial intelligence

Pokee AI推出零代码AI Agent平台，并开源核心模型

简报：

• 初创平台Pokee AI推出新工具，用户仅需输入一句自然语言指令，即可创建和执行AI智能工作流，无需编码。
• 该平台通过自研的“提示到工作流”引擎自动拆解任务，内置基于强化学习的智能代理，并已集成Google Workspace、Slack、Notion等数千款主流应用。
• 公司成立于2024年，已获得高通创投、三星NEXT等投资的1200万美元种子轮融资。
• Pokee AI已将其核心模型PokeeResearch-7B（70亿参数）在Hugging Face上开源，供开发者下载使用。

相关链接：

• 一句话生成AI Agent！Pokee AI零代码引爆自动化革命

Pokee AI Launches No-Code AI Agent Platform and Open-Sources Core Model

Brief:

• Startup platform Pokee AI has launched a new tool that allows users to create and execute AI-powered smart workflows with just a single natural language instruction, no coding required. 🚀
• The platform automatically breaks down tasks using its self-developed 'Prompt-to-Workflow' engine, features built-in reinforcement learning-based intelligent agents, and has integrated thousands of popular applications such as Google Workspace, Slack, and Notion. 💡
• Founded in 2024, the company has secured dJyH6kDkfC5RlM16.answerText2 million in seed funding from investors including Qualcomm Ventures and Samsung NEXT.
• Pokee AI has open-sourced its core model, PokeeResearch-7B (7 billion parameters), on Hugging Face for developers to download and use. 🤖

Related Links:

• Generate AI Agents with One Sentence! Pokee AI's No-Code Platform Ignites an Automation Revolution

谷歌携手信实Jio，欲借AI普惠策略抢占印度市场

简报：谷歌宣布与印度最大电信运营商信实Jio达成合作，将向其超过5亿用户免费提供Gemini人工智能服务，首批面向18至25岁年轻用户，提供包括Gemini 2.5 Pro、NotebookLM工具及2TB云存储在内，价值约396美元的AI权益。

相关链接：

• 谷歌联手印度信实Jio，免费向5亿用户推送Gemini AI！人均396美元福利引爆新兴市场
• Gemini 3明年登场！谷歌全力追赶GPT-5，6.5亿用户成AI反攻关键筹码
• 谷歌启动高级搜索广告实验，未来将在其AI生成式内容中展示广告
• Siri要翻身了？苹果明年三月推“Apple Intelligence”大改版，竟用上谷歌Gemini！

Google Teams Up with Reliance Jio, Aims to Capture Indian Market with Inclusive AI Strategy 🚀

Brief: Google announced a partnership with Reliance Jio, India's largest telecom operator, to offer its Gemini AI services for free to over 500 million users. Initially targeting young users aged 18 to 25, this initiative provides AI benefits worth approximately $396, including Gemini 2.5 Pro, NotebookLM tools, and 2TB of cloud storage. 💡🇮🇳

Related Links:

• Google Teams Up with India's Reliance Jio to Offer Free Gemini AI to 500 Million Users! $396 Per Person Benefit Ignites Emerging Market
• Gemini 3 Arriving Next Year! Google Goes All Out to Catch Up with GPT-5, 650 Million Users Become Key Bargaining Chip for AI Counterattack
• Google Launches Advanced Search Ad Experiment, Will Display Ads in its AI-Generated Content in the Future
• Is Siri Making a Comeback? Apple to Launch Major "Apple Intelligence" Revamp Next March, Reportedly Using Google Gemini!

ChatGPT获重大安全升级，可提前识别自杀等风险信号

简报：

• OpenAI 宣布对 ChatGPT 语言模型进行全面升级，旨在提早识别用户的心理危机、自残及自杀等风险话题。
• 本次升级是 OpenAI 与超过170名心理学、精神医学及危机干预领域专家合作的成果，能捕捉更细微的情感困扰信号。
• 内部测试显示，基于 GPT-5 的新模型可减少80%的不恰当回应，并在长时间复杂情绪对话中保持95%以上的可靠性。
• 模型升级后，将通过温和的方式引导用户寻求真实世界的专业帮助，避免用户陷入更大压力。

相关链接：

• OpenAI 改进 ChatGPT 语言模型，提早识别心理压力与风险话题
• OpenAI 升级 ChatGPT：提前识别心理危机，减少80%不当回应

ChatGPT Receives Major Safety Upgrade, Can Proactively Identify Risk Signals Like Suicide 🛡️

Brief:

• OpenAI has announced a comprehensive upgrade to its ChatGPT language model, aimed at proactively identifying user mental health crises, self-harm, and suicidal ideation, among other risk topics.
• This upgrade is the result of OpenAI's collaboration with over 170 experts in psychology, psychiatry, and crisis intervention, enabling it to detect more subtle signals of emotional distress. 🧠
• Internal tests show that the new GPT-5-based model can reduce inappropriate responses by 80% and maintain over 95% reliability in prolonged, complex emotional conversations.
• Following the upgrade, the model will gently guide users to seek real-world professional help, preventing them from falling into greater distress. 🙏

Related Links:

• OpenAI Improves ChatGPT Language Model to Proactively Identify Mental Stress and Risk Topics
• OpenAI Upgrades ChatGPT: Proactively Identifies Mental Crises, Reduces Inappropriate Responses by 80%

库克确认苹果AI战略：开放收购并与谷歌OpenAI合作

简报： 苹果公司CEO蒂姆·库克在2025财年第四季度财报会议上表示，公司对人工智能领域的收购与合作持开放态度，将通过自研基础模型、与谷歌和OpenAI等第三方合作以及收购的三重策略，推进代号为“Apple Intelligence”的AI计划，并预计在2026年发布新版Siri。

相关链接：

• Siri要翻身了？苹果明年三月推“Apple Intelligence”大改版，竟用上谷歌Gemini！
• 库克:苹果将在人工智能领域寻求更多收购与合作机会

Cook Confirms Apple's AI Strategy: Open to Acquisitions and Collaborations with Google, OpenAI

Brief: Apple CEO Tim Cook stated during the Q4 FY2025 earnings call that the company is open to acquisitions and collaborations in the artificial intelligence sector. 🤝 Apple will advance its AI initiative, codenamed "Apple Intelligence," through a three-pronged strategy: developing its own foundational models, partnering with third parties like Google and OpenAI, and making acquisitions. 🍎 The company anticipates releasing a new version of Siri in 2026. 🚀

Related Links:

• Is Siri getting a makeover? Apple to launch "Apple Intelligence" major revamp next March, reportedly using Google Gemini!
• Cook: Apple to seek more acquisition and collaboration opportunities in AI

美团开源全模态大模型，支持实时音视频交互达SOTA水平

简报：

• 美团于2025年11月3日发布并开源了其LongCat-Flash系列的新成员——LongCat-Flash-Omni全模态大模型。
• 该模型是业界首个实现“全模态覆盖、端到端架构、大参数量高效推理”于一体的开源模型，可支持毫秒级响应的实时音视频交互。
• 综合评估结果显示，LongCat-Flash-Omni在全模态基准测试中达到开源最先进（SOTA）水平，在文本、图像、音视频理解及语音感知与生成等单模态任务中也表现出竞争力。
• 与此同时，集成了该模型技术的官方应用“LongCat”已正式上线，支持联网搜索和语音通话功能。
• 模型代码和权重已在Hugging Face和Github上同步开源。

相关链接：

• 美团发布并开源 LongCat-Flash-Omni 模型：支持实时音视频交互，达到 SOTA 水平
• 美团 LongCat 大模型 App 正式上线！语音通话与联网搜索让你与 AI 更亲密
• Hugging Face 开源地址
• Github 开源地址

Meituan Open-Sources Omnimodal Large Model, Achieving SOTA Level in Real-time Audio-Video Interaction

Brief:

• Meituan released and open-sourced LongCat-Flash-Omni, a new member of its LongCat-Flash series and an omnimodal large model, on November 3, 2025.
• This model is the industry's first open-source model to integrate "full modality coverage, end-to-end architecture, and efficient inference with large parameters," capable of supporting real-time audio and video interactions with millisecond-level response.
• Comprehensive evaluation results indicate that LongCat-Flash-Omni achieves state-of-the-art (SOTA) performance in open-source omnimodal benchmarks, while also demonstrating competitive performance in single-modal tasks such as text, image, audio-video understanding, and speech perception and generation. 🚀
• Concurrently, the official application "LongCat," which integrates this model's technology, has been officially launched, supporting internet search and voice call functionalities. 📱
• The model code and weights have been simultaneously open-sourced on Hugging Face and Github. 💡

Related Links:

• Meituan Releases and Open-Sources LongCat-Flash-Omni Model: Supports Real-time Audio and Video Interaction, Achieves SOTA Level
• Meituan LongCat Large Model App Officially Launched! Voice Calls and Internet Search Bring You Closer to AI
• Hugging Face Open-Source Link
• Github Open-Source Link